Many resources are needed to download a project. Please understand that we have to compensate our server costs. Thank you in advance.
Project price only 1 $
You can buy this project and download/modify it how often you want.
org.nd4j.linalg.cpu.nativecpu.CpuNDArrayFactory Maven / Gradle / Ivy
/*******************************************************************************
* Copyright (c) 2015-2018 Skymind, Inc.
*
* This program and the accompanying materials are made available under the
* terms of the Apache License, Version 2.0 which is available at
* https://www.apache.org/licenses/LICENSE-2.0.
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS, WITHOUT
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the
* License for the specific language governing permissions and limitations
* under the License.
*
* SPDX-License-Identifier: Apache-2.0
******************************************************************************/
package org.nd4j.linalg.cpu.nativecpu;
import lombok.extern.slf4j.Slf4j;
import lombok.val;
import org.nd4j.config.ND4JSystemProperties;
import org.nd4j.linalg.api.buffer.LongBuffer;
import org.nd4j.linalg.api.ops.performance.PerformanceTracker;
import org.nd4j.linalg.api.shape.options.ArrayOptionsHelper;
import org.nd4j.linalg.api.shape.options.ArrayType;
import org.nd4j.linalg.compression.CompressionUtils;
import org.nd4j.linalg.memory.MemcpyDirection;
import org.nd4j.linalg.primitives.Pair;
import org.bytedeco.javacpp.*;
import org.nd4j.linalg.api.buffer.DataBuffer;
import org.nd4j.linalg.api.memory.MemoryWorkspace;
import org.nd4j.linalg.api.ndarray.INDArray;
import org.nd4j.linalg.api.shape.Shape;
import org.nd4j.linalg.api.shape.options.ArrayOptionsHelper;
import org.nd4j.linalg.api.shape.options.ArrayType;
import org.nd4j.linalg.cache.TADManager;
import org.nd4j.linalg.compression.CompressedDataBuffer;
import org.nd4j.linalg.compression.CompressionDescriptor;
import org.nd4j.linalg.compression.CompressionType;
import org.nd4j.linalg.compression.CompressionUtils;
import org.nd4j.linalg.cpu.nativecpu.blas.*;
import org.nd4j.linalg.exception.ND4JIllegalStateException;
import org.nd4j.linalg.factory.Nd4j;
import org.nd4j.linalg.primitives.Pair;
import org.nd4j.linalg.util.ArrayUtil;
import org.nd4j.nativeblas.BaseNativeNDArrayFactory;
import org.nd4j.nativeblas.LongPointerWrapper;
import org.nd4j.nativeblas.NativeOpsHolder;
import java.util.*;
/**
* {@link org.nd4j.linalg.factory.NDArrayFactory}
* for cpus and the nd4j-native backend.
*
* @author Adam Gibson
*/
@Slf4j
public class CpuNDArrayFactory extends BaseNativeNDArrayFactory {
protected ThreadLocal extrazA = new ThreadLocal<>();
protected ThreadLocal extrazB = new ThreadLocal<>();
protected ThreadLocal extrazSize = new ThreadLocal<>();
public CpuNDArrayFactory() {}
static {
//invoke the override
Nd4j.getBlasWrapper();
}
public CpuNDArrayFactory(DataBuffer.Type dtype, Character order) {
super(dtype, order);
}
public CpuNDArrayFactory(DataBuffer.Type dtype, char order) {
super(dtype, order);
}
@Override
public void createBlas() {
String lib = System.getProperty(ND4JSystemProperties.ND4J_CPU_LOAD_OPENBLAS,
System.getProperty(ND4JSystemProperties.ND4J_CPU_LOAD_OPENBLAS_NOLAPACK, "")).toLowerCase();
if (lib.trim().length() == 0) {
// try to load by default the LAPACK-less version of MKL bundled with MKL-DNN
System.setProperty(ND4JSystemProperties.ND4J_CPU_LOAD_OPENBLAS_NOLAPACK, "mklml");
}
blas = new CpuBlas();
// TODO: add batched gemm here
PointerPointer functions = new PointerPointer(10);
functions.put(0, Loader.addressof("cblas_sgemv"));
functions.put(1, Loader.addressof("cblas_dgemv"));
functions.put(2, Loader.addressof("cblas_sgemm"));
functions.put(3, Loader.addressof("cblas_dgemm"));
functions.put(4, Loader.addressof("cblas_sgemm_batch"));
functions.put(5, Loader.addressof("cblas_dgemm_batch"));
functions.put(6, Loader.addressof("LAPACKE_sgesvd"));
functions.put(7, Loader.addressof("LAPACKE_dgesvd"));
functions.put(8, Loader.addressof("LAPACKE_sgesdd"));
functions.put(9, Loader.addressof("LAPACKE_dgesdd"));
nativeOps.initializeFunctions(functions);
}
@Override
public void createLevel1() {
level1 = new CpuLevel1();
}
@Override
public void createLevel2() {
level2 = new CpuLevel2();
}
@Override
public void createLevel3() {
level3 = new CpuLevel3();
}
@Override
public void createLapack() {
lapack = new CpuLapack();
}
@Override
public INDArray create(int[] shape, DataBuffer buffer) {
return new NDArray(shape, buffer);
}
/**
* Create an ndarray with the given data layout
*
* @param data the data to create the ndarray with
* @return the ndarray with the given data layout
*/
@Override
public INDArray create(double[][] data) {
return new NDArray(data);
}
@Override
public INDArray create(double[][] data, char ordering) {
return new NDArray(data, ordering);
}
@Override
public INDArray create(DataBuffer data) {
return new NDArray(data);
}
@Override
public INDArray create(DataBuffer data, long rows, long columns, int[] stride, long offset) {
return create(data, new long[]{rows, columns}, ArrayUtil.toLongArray(stride), offset);
}
@Override
public INDArray create(long rows, long columns, long[] stride, long offset) {
return create(new long[]{rows, columns}, stride, offset);
}
@Override
public INDArray create(int[] shape, char ordering) {
return new NDArray(shape, Nd4j.getStrides(shape, ordering), 0, ordering);
}
@Override
public INDArray create(long[] shape, char ordering) {
return new NDArray(shape, Nd4j.getStrides(shape, ordering), 0, ordering);
}
@Override
public INDArray createUninitialized(int[] shape, char ordering) {
return new NDArray(shape, Nd4j.getStrides(shape, ordering), 0, ordering, false);
}
@Override
public INDArray createUninitialized(long[] shape, char ordering) {
return new NDArray(shape, Nd4j.getStrides(shape, ordering), 0, ordering, false);
}
@Override
public INDArray createUninitializedDetached(int[] shape, char ordering) {
MemoryWorkspace workspace = Nd4j.getMemoryManager().getCurrentWorkspace();
Nd4j.getMemoryManager().setCurrentWorkspace(null);
INDArray ret = new NDArray(shape, Nd4j.getStrides(shape, ordering), 0, ordering, false);
Nd4j.getMemoryManager().setCurrentWorkspace(workspace);
return ret;
}
@Override
public INDArray createUninitializedDetached(long[] shape, char ordering) {
MemoryWorkspace workspace = Nd4j.getMemoryManager().getCurrentWorkspace();
Nd4j.getMemoryManager().setCurrentWorkspace(null);
INDArray ret = new NDArray(shape, Nd4j.getStrides(shape, ordering), 0, ordering, false);
Nd4j.getMemoryManager().setCurrentWorkspace(workspace);
return ret;
}
@Override
public INDArray create(DataBuffer data, int[] newShape, int[] newStride, long offset, char ordering) {
return new NDArray(data, newShape, newStride, offset, ordering);
}
@Override
public INDArray create(float[] data, int[] shape, long offset, Character order) {
return new NDArray(data, shape, offset, order);
}
@Override
public INDArray create(float[] data, long[] shape, long offset, Character order) {
return new NDArray(data, shape, offset, order);
}
@Override
public INDArray create(float[] data, long rows, long columns, int[] stride, long offset, char ordering) {
return create(data, new long[]{rows, columns}, ArrayUtil.toLongArray(stride), offset, ordering);
}
@Override
public INDArray create(double[] data, int[] shape, char ordering) {
return new NDArray(Nd4j.createBuffer(data), shape, ordering);
}
@Override
public INDArray create(double[] data, long[] shape, char ordering) {
return create(data, shape, (Character) ordering);
}
@Override
public INDArray create(float[] data, long[] shape, char ordering) {
return create(data, shape, (Character) ordering);
}
@Override
public INDArray create(List list, int[] shape, char ordering) {
return new NDArray(list, shape, ordering);
}
@Override
public INDArray create(List list, long[] shape, char ordering) {
return new NDArray(list, shape, ordering);
}
@Override
public INDArray create(double[] data, int[] shape, long offset) {
return new NDArray(Nd4j.createBuffer(data), shape, offset);
}
@Override
public INDArray create(double[] data, long[] shape, long offset, Character order) {
return new NDArray(data, shape, offset, order.charValue());
}
@Override
public INDArray create(double[] data, int[] shape, int[] stride, long offset, char ordering) {
return new NDArray(Nd4j.createBuffer(data), shape, stride, offset, ordering);
}
@Override
public INDArray create(double[] data, long[] shape, long[] stride, long offset, char ordering) {
return new NDArray(Nd4j.createBuffer(data), shape, stride, offset, ordering);
}
@Override
public INDArray create(float[] data, long[] shape, long[] stride, long offset, char ordering) {
return new NDArray(Nd4j.createBuffer(data), shape, stride, offset, ordering);
}
@Override
public INDArray create(float[] data, long[] shape, long[] stride, long offset) {
return new NDArray(data, shape, stride, offset, Nd4j.order());
}
@Override
public INDArray create(double[] data, long[] shape, long[] stride, long offset) {
return new NDArray(data, shape, stride, offset, Nd4j.order());
}
@Override
public INDArray create(DataBuffer data, long[] shape) {
return new NDArray(data, shape);
}
@Override
public INDArray create(DataBuffer data, long[] shape, long[] stride, long offset) {
return create(data, shape, stride, offset, Nd4j.order());
}
@Override
public INDArray create(DataBuffer data, long[] shape, long[] stride, long offset, char ordering) {
return new NDArray(data, shape, stride, offset, ordering);
}
@Override
public INDArray create(float[] data, long[] shape, long[] stride, char order, long offset) {
return new NDArray(data, shape, stride, offset, order);
}
/**
* Creates an ndarray with the specified shape
*
* @param data
* @param shape the shape of the ndarray
* @param stride the stride for the ndarray
* @param offset the offset of the ndarray
* @return the instance
*/
@Override
public INDArray create(float[] data, int[] shape, int[] stride, long offset) {
return new NDArray(data, shape, stride, offset);
}
/**
* Creates an ndarray with the specified shape
*
* @param data
* @param shape the shape of the ndarray
* @param stride the stride for the ndarray
* @param offset the offset of the ndarray
* @return the instance
*/
@Override
public INDArray create(double[] data, int[] shape, int[] stride, long offset) {
return new NDArray(data, shape, stride, offset);
}
@Override
public INDArray create(DataBuffer data, int[] shape) {
return new NDArray(data, shape);
}
@Override
public INDArray create(DataBuffer data, int[] shape, int[] stride, long offset) {
return new NDArray(data, shape, stride, offset, Nd4j.order());
}
/**
* Creates an ndarray with the specified shape
*
* @param list
* @param shape the shape of the ndarray
* @return the instance
*/
@Override
public INDArray create(List list, int[] shape) {
return new NDArray(list, shape, Nd4j.getStrides(shape));
}
@Override
public INDArray create(List list, long[] shape) {
return new NDArray(list, shape, Nd4j.getStrides(shape));
}
@Override
public INDArray empty(DataBuffer.Type type) {
long extras = ArrayOptionsHelper.setOptionBit(0L, ArrayType.EMPTY);
extras = ArrayOptionsHelper.setOptionBit(extras, type);
val shape = Nd4j.getShapeInfoProvider().createShapeInformation(new int[0], new int[0],0,1,'c', extras);
return new NDArray(null, (LongBuffer) shape.getFirst(), shape.getSecond());
}
@Override
public INDArray create(float[][] floats) {
return new NDArray(floats);
}
@Override
public INDArray create(float[][] data, char ordering) {
return new NDArray(data, ordering);
}
@Override
public INDArray create(float[] data, int[] shape, int[] stride, long offset, char ordering) {
return new NDArray(data, shape, stride, offset, ordering);
}
@Override
public INDArray create(DataBuffer buffer, int[] shape, long offset) {
return new NDArray(buffer, shape, Nd4j.getStrides(shape), offset);
}
@Override
public INDArray create(float[] data, int[] shape, long offset) {
return new NDArray(data, shape, offset);
}
@Override
public INDArray toFlattened(char order, Collection matrices) {
int length = 0;
for (INDArray m : matrices)
length += m.length();
INDArray ret = Nd4j.create(new int[] {1, length}, order);
int linearIndex = 0;
PointerPointer dummy = new PointerPointer(new Pointer[] {null});
for (INDArray m : matrices) {
Nd4j.getCompressor().autoDecompress(m);
if (m.ordering() == order && m.data().allocationMode() == DataBuffer.AllocationMode.HEAP
&& Shape.strideDescendingCAscendingF(m) && Shape.isContiguousInBuffer(m)) {
//Can do array copy
int retFrom = linearIndex;
long mFrom = m.offset();
Object arr = m.data().array();
if (arr instanceof float[]) {
float[] mData = (float[]) arr;
float[] retData = (float[]) ret.data().array();
// FIXME: LONG
// FIXME: int cast
System.arraycopy(mData, (int) mFrom, retData, retFrom, (int) m.length());
} else {
double[] mData = (double[]) arr;
double[] retData = (double[]) ret.data().array();
// FIXME: LONG
// FIXME: int cast
System.arraycopy(mData, (int) mFrom, retData, retFrom, (int) m.length());
}
linearIndex += m.length();
} else {
if (m.data().dataType() == DataBuffer.Type.DOUBLE) {
nativeOps.flattenDouble(dummy, linearIndex, order, (DoublePointer) ret.data().addressPointer(),
(LongPointer) ret.shapeInfoDataBuffer().addressPointer(),
(DoublePointer) m.data().addressPointer(),
(LongPointer) m.shapeInfoDataBuffer().addressPointer());
} else if (m.data().dataType() == DataBuffer.Type.FLOAT) {
nativeOps.flattenFloat(dummy, linearIndex, order, (FloatPointer) ret.data().addressPointer(),
(LongPointer) ret.shapeInfoDataBuffer().addressPointer(),
(FloatPointer) m.data().addressPointer(),
(LongPointer) m.shapeInfoDataBuffer().addressPointer());
} else {
throw new UnsupportedOperationException("Illegal data opType for copy");
}
//Works for all cases...
/* NdIndexIterator iter = new NdIndexIterator(order, m.shape());
while (iter.hasNext()) {
ret.putScalar(linearIndex++, m.getDouble(iter.next()));
}*/
linearIndex += m.length();
}
}
return ret;
}
@Override
public INDArray[] tear(INDArray tensor, int... dimensions) {
if (tensor.isCompressed())
Nd4j.getCompressor().decompressi(tensor);
Arrays.sort(dimensions);
Pair tadBuffers = Nd4j.getExecutioner().getTADManager().getTADOnlyShapeInfo(tensor, dimensions);
long tadLength = 1;
long[] shape = new long[dimensions.length];
for (int i = 0; i < dimensions.length; i++) {
tadLength *= tensor.shape()[dimensions[i]];
shape[i] = tensor.shape()[dimensions[i]];
}
int numTads = (int)(tensor.lengthLong() / tadLength);
INDArray[] result = new INDArray[numTads];
PointerPointer targets = new PointerPointer(numTads);
for (int x = 0; x < numTads; x++) {
result[x] = Nd4j.createUninitialized(shape);
targets.put(x, result[x].data().pointer());
}
if (Nd4j.dataType() == DataBuffer.Type.DOUBLE) {
nativeOps.tearDouble(null,
(DoublePointer) tensor.data().pointer(),
(LongPointer) tensor.shapeInfoDataBuffer().pointer(),
targets,
(LongPointer) result[0].shapeInfoDataBuffer().pointer(),
(LongPointer) tadBuffers.getFirst().pointer(),
new LongPointerWrapper(tadBuffers.getSecond().pointer())
);
} else if (Nd4j.dataType() == DataBuffer.Type.FLOAT) {
nativeOps.tearFloat(null,
(FloatPointer) tensor.data().pointer(),
(LongPointer) tensor.shapeInfoDataBuffer().pointer(),
targets,
(LongPointer) result[0].shapeInfoDataBuffer().pointer(),
(LongPointer) tadBuffers.getFirst().pointer(),
new LongPointerWrapper(tadBuffers.getSecond().pointer())
);
} else if (Nd4j.dataType() == DataBuffer.Type.HALF) {
throw new UnsupportedOperationException("Half precision isn't supported for CPU backend");
}
return result;
}
/**
* concatenate ndarrays along a dimension
*
* @param dimension the dimension to concatenate along
* @param toConcat the ndarrays to concatenate
* @return the concatenate ndarrays
*/
@Override
public INDArray concat(int dimension, INDArray... toConcat) {
if (toConcat == null || toConcat.length == 0)
throw new ND4JIllegalStateException("Can't concatenate 0 arrays");
if (toConcat.length == 1)
return toConcat[0];
// if reusable var wasn't created for this thread, or is smaller then needed - set it to new value
if (extrazA.get() == null || extrazB.get() == null || extrazSize.get() == null || extrazSize.get() < toConcat.length) {
extrazA.set(new PointerPointer(toConcat.length));
extrazB.set(new PointerPointer(toConcat.length));
extrazSize.set(toConcat.length);
}
PointerPointer shapeInfoPointers = extrazA.get();
PointerPointer dataPointers = extrazB.get();
int sumAlongDim = 0;
long[] outputShape = ArrayUtil.copy(toConcat[0].shape());
for (int i = 0; i < toConcat.length; i++) {
if (toConcat[i].isCompressed())
Nd4j.getCompressor().decompressi(toConcat[i]);
shapeInfoPointers.put(i, toConcat[i].shapeInfoDataBuffer().addressPointer());
dataPointers.put(i, toConcat[i].data().addressPointer());
sumAlongDim += toConcat[i].size(dimension);
for (int j = 0; j < toConcat[i].rank(); j++)
if (j != dimension && toConcat[i].size(j) != outputShape[j]) {
throw new IllegalArgumentException(
"Illegal concatenation at array " + i + " and shape element " + j);
}
//log.info("Shape[{}]: {}", i, Arrays.toString(toConcat[i].shapeInfoDataBuffer().asInt()));
}
outputShape[dimension] = sumAlongDim;
//PointerPointer dummy = new PointerPointer(new Pointer[] {null});
INDArray ret = Nd4j.createUninitialized(outputShape, Nd4j.order());
if (ret.data().dataType() == DataBuffer.Type.DOUBLE) {
nativeOps.concatDouble(null, dimension, toConcat.length, dataPointers, shapeInfoPointers,
(DoublePointer) ret.data().addressPointer(),
(LongPointer) ret.shapeInfoDataBuffer().addressPointer(),
//new PointerPointer(new Pointer[] {null}), new PointerPointer(new Pointer[] {null}));
null, null);
} else if (ret.data().dataType() == DataBuffer.Type.FLOAT) {
nativeOps.concatFloat(null, dimension, toConcat.length, dataPointers, shapeInfoPointers,
(FloatPointer) ret.data().addressPointer(),
(LongPointer) ret.shapeInfoDataBuffer().addressPointer(),
//new PointerPointer(new Pointer[] {null}), new PointerPointer(new Pointer[] {null}));
null, null);
} else if (ret.data().dataType() == DataBuffer.Type.HALF) {
nativeOps.concatHalf(null, dimension, toConcat.length, dataPointers, shapeInfoPointers,
(ShortPointer) ret.data().addressPointer(),
(LongPointer) ret.shapeInfoDataBuffer().addressPointer(),
//new PointerPointer(new Pointer[]{null}), new PointerPointer(new Pointer[]{null}));
null, null);
} else {
throw new ND4JIllegalStateException("Unknown dataType: " + ret.data().dataType());
}
return ret;
// return super.concat(dimension,toConcat);
}
/**
* For CPU backend this method is equal to concat()
*
* @param dimension the dimension to concatneate along
* @param toConcat the ndarrays to concateneate
* @return
*/
@Override
public INDArray specialConcat(int dimension, INDArray... toConcat) {
return concat(dimension, toConcat);
}
/**
* This method produces concatenated array, that consist from tensors, fetched from source array, against some dimension and specified indexes
*
* @param source source tensor
* @param sourceDimension dimension of source tensor
* @param indexes indexes from source array
* @return
*/
@Override
public INDArray pullRows(INDArray source, int sourceDimension, int[] indexes) {
return pullRows(source, sourceDimension, ArrayUtil.toLongArray(indexes));
}
@Override
public INDArray pullRows(INDArray source, int sourceDimension, long[] indexes) {
return pullRows(source, sourceDimension, indexes, Nd4j.order());
}
/**
* This method produces concatenated array, that consist from tensors, fetched from source array, against some dimension and specified indexes
*
* @param source source tensor
* @param sourceDimension dimension of source tensor
* @param indexes indexes from source array
* @return
*/
public INDArray pullRows(INDArray source, int sourceDimension, long[] indexes, char order) {
if (indexes == null || indexes.length < 1)
throw new IllegalStateException("Indexes can't be null or zero-length");
long[] shape;
if (sourceDimension == 1)
shape = new long[] {indexes.length, source.shape()[sourceDimension]};
else if (sourceDimension == 0)
shape = new long[] {source.shape()[sourceDimension], indexes.length};
else
throw new UnsupportedOperationException("2D input is expected");
return pullRows(source, Nd4j.createUninitialized(shape, order), sourceDimension, indexes);
}
@Override
public INDArray pullRows(INDArray source, int sourceDimension, int[] indexes, char order) {
return pullRows(source, sourceDimension, ArrayUtil.toLongArray(indexes), order);
}
@Override
public INDArray pullRows(INDArray source, INDArray destination, int sourceDimension, int[] indexes) {
return pullRows(source, destination, sourceDimension, ArrayUtil.toLongArray(indexes));
}
public INDArray pullRows(INDArray source, INDArray destination, int sourceDimension, long[] indexes) {
if (indexes == null || indexes.length < 1)
throw new IllegalStateException("Indexes can't be null or zero-length");
long[] shape = null;
if (sourceDimension == 1)
shape = new long[] {indexes.length, source.shape()[sourceDimension]};
else if (sourceDimension == 0)
shape = new long[] {source.shape()[sourceDimension], indexes.length};
else
throw new UnsupportedOperationException("2D input is expected");
INDArray ret = destination;
if(ret == null){
ret = Nd4j.createUninitialized(shape, order);
} else {
if(!Arrays.equals(shape, destination.shape())){
throw new IllegalStateException("Cannot pull rows into destination array: expected destination array of" +
" shape " + Arrays.toString(shape) + " but got destination array of shape " + Arrays.toString(destination.shape()));
}
}
Nd4j.getCompressor().autoDecompress(source);
PointerPointer dummy = new PointerPointer(new Pointer[] {null});
TADManager tadManager = Nd4j.getExecutioner().getTADManager();
Pair tadBuffers = tadManager.getTADOnlyShapeInfo(source, new int[] {sourceDimension});
Pair zTadBuffers = tadManager.getTADOnlyShapeInfo(ret, new int[] {sourceDimension});
Pointer hostTadShapeInfo = tadBuffers.getFirst().addressPointer();
Pointer zTadShapeInfo = zTadBuffers.getFirst().addressPointer();
LongPointer pIndex = new LongPointer(indexes);
DataBuffer offsets = tadBuffers.getSecond();
Pointer hostTadOffsets = offsets == null ? null : offsets.addressPointer();
DataBuffer zOffsets = zTadBuffers.getSecond();
Pointer zTadOffsets = zOffsets == null ? null : zOffsets.addressPointer();
if (ret.data().dataType() == DataBuffer.Type.DOUBLE) {
nativeOps.pullRowsDouble(dummy, (DoublePointer) source.data().addressPointer(),
(LongPointer) source.shapeInfoDataBuffer().addressPointer(),
(DoublePointer) ret.data().addressPointer(),
(LongPointer) ret.shapeInfoDataBuffer().addressPointer(), indexes.length, pIndex,
(LongPointer) hostTadShapeInfo, new LongPointerWrapper(hostTadOffsets), (LongPointer) zTadShapeInfo,
new LongPointerWrapper(zTadOffsets));
} else if (ret.data().dataType() == DataBuffer.Type.FLOAT) {
nativeOps.pullRowsFloat(dummy, (FloatPointer) source.data().addressPointer(),
(LongPointer) source.shapeInfoDataBuffer().addressPointer(),
(FloatPointer) ret.data().addressPointer(),
(LongPointer) ret.shapeInfoDataBuffer().addressPointer(), indexes.length, pIndex,
(LongPointer) hostTadShapeInfo, new LongPointerWrapper(hostTadOffsets), (LongPointer) zTadShapeInfo,
new LongPointerWrapper(zTadOffsets));
} else {
nativeOps.pullRowsHalf(dummy, (ShortPointer) source.data().addressPointer(),
(LongPointer) source.shapeInfoDataBuffer().addressPointer(),
(ShortPointer) ret.data().addressPointer(),
(LongPointer) ret.shapeInfoDataBuffer().addressPointer(), indexes.length, pIndex,
(LongPointer) hostTadShapeInfo, new LongPointerWrapper(hostTadOffsets), (LongPointer) zTadShapeInfo,
new LongPointerWrapper(zTadOffsets));
}
return ret;
}
public INDArray accumulate(INDArray target, INDArray... arrays) {
if (arrays == null || arrays.length == 0)
throw new RuntimeException("Input arrays are missing");
if (arrays.length == 1)
return target.addi(arrays[0]);
long len = target.lengthLong();
PointerPointer dataPointers = new PointerPointer(arrays.length);
for (int i = 0; i < arrays.length; i++) {
Nd4j.getCompressor().autoDecompress(arrays[i]);
if (arrays[i].elementWiseStride() != 1)
throw new ND4JIllegalStateException("Native accumulation is applicable only to continuous INDArrays");
if (arrays[i].lengthLong() != len)
throw new ND4JIllegalStateException("All arrays should have equal length for accumulation");
dataPointers.put(i, arrays[i].data().addressPointer());
}
if (target.data().dataType() == DataBuffer.Type.DOUBLE) {
nativeOps.accumulateDouble(null, dataPointers, (DoublePointer) target.data().addressPointer(), arrays.length, len);
} else if (target.data().dataType() == DataBuffer.Type.FLOAT) {
nativeOps.accumulateFloat(null, dataPointers, (FloatPointer) target.data().addressPointer(), arrays.length, len);
} else {
nativeOps.accumulateHalf(null, dataPointers, (ShortPointer) target.data().addressPointer(), arrays.length, len);
}
return target;
}
/**
* This method averages input arrays, and returns averaged array
*
* @param target
* @param arrays
* @return
*/
@Override
public INDArray average(INDArray target, INDArray[] arrays) {
if (arrays == null || arrays.length == 0)
throw new RuntimeException("Input arrays are missing");
if (arrays.length == 1)
return target.assign(arrays[0]);
long len = target != null ? target.lengthLong() : arrays[0].length();
PointerPointer dataPointers = new PointerPointer(arrays.length);
for (int i = 0; i < arrays.length; i++) {
Nd4j.getCompressor().autoDecompress(arrays[i]);
if (arrays[i].elementWiseStride() != 1)
throw new ND4JIllegalStateException("Native averaging is applicable only to continuous INDArrays");
if (arrays[i].lengthLong() != len)
throw new ND4JIllegalStateException("All arrays should have equal length for averaging");
dataPointers.put(i, arrays[i].data().addressPointer());
}
if (arrays[0].data().dataType() == DataBuffer.Type.DOUBLE) {
nativeOps.averageDouble(null, dataPointers, target == null ? null : (DoublePointer) target.data().addressPointer(), arrays.length,
len, true);
} else if (arrays[0].data().dataType() == DataBuffer.Type.FLOAT) {
nativeOps.averageFloat(null, dataPointers, target == null ? null : (FloatPointer) target.data().addressPointer(), arrays.length,
len, true);
} else {
nativeOps.averageHalf(null, dataPointers, target == null ? null : (ShortPointer) target.data().addressPointer(), arrays.length, len, true);
}
return target;
}
/**
* This method averages input arrays, and returns averaged array
*
* @param target
* @param arrays
* @return
*/
@Override
public INDArray average(INDArray target, Collection arrays) {
return average(target, arrays.toArray(new INDArray[0]));
}
@Override
public INDArray average(INDArray[] arrays) {
if (arrays == null || arrays.length == 0)
throw new RuntimeException("Input arrays are missing");
INDArray ret = Nd4j.createUninitialized(arrays[0].shape(), arrays[0].ordering());
return average(ret, arrays);
}
@Override
public INDArray average(Collection arrays) {
return average(arrays.toArray(new INDArray[0]));
}
/**
* In place shuffle of an ndarray
* along a specified set of dimensions
*
* @param array the ndarray to shuffle
* @param dimension the dimension to do the shuffle
* @return
*/
@Override
public void shuffle(INDArray array, Random rnd, int... dimension) {
shuffle(Collections.singletonList(array), rnd, dimension);
}
/**
* Symmetric in place shuffle of an ndarray
* along a specified set of dimensions. All arrays
*
* @param array the ndarray to shuffle
* @param dimension the dimension to do the shuffle
* @return
*/
@Override
public void shuffle(Collection array, Random rnd, int... dimension) {
shuffle(new ArrayList(array), rnd, Collections.singletonList(dimension));
}
/**
* Symmetric in place shuffle of an ndarray
* along a specified set of dimensions. Each array in list should have it's own dimension at the same index of dimensions array
*
* @param arrays the ndarrays to shuffle
* @param dimensions the dimensions to do the shuffle
* @return
*/
@Override
public void shuffle(List arrays, Random rnd, List dimensions) {
if (dimensions == null || dimensions.size() == 0)
throw new RuntimeException("Dimension can't be null or 0-length");
if (arrays == null || arrays.size() == 0)
throw new RuntimeException("No input arrays provided");
if (dimensions.size() > 1 && arrays.size() != dimensions.size())
throw new IllegalStateException("Number of dimensions do not match number of arrays to shuffle");
int tadLength = 1;
for (int i = 0; i < dimensions.get(0).length; i++) {
tadLength *= arrays.get(0).shape()[dimensions.get(0)[i]];
}
long numTads = arrays.get(0).length() / tadLength;
val map = ArrayUtil.buildInterleavedVector(rnd, (int) numTads);
PointerPointer dataPointers = new PointerPointer(arrays.size());
PointerPointer shapePointers = new PointerPointer(arrays.size());
PointerPointer tadPointers = new PointerPointer(arrays.size());
PointerPointer offsetPointers = new PointerPointer(arrays.size());
PointerPointer dummy = new PointerPointer(new Pointer[] {null});
List> list = new ArrayList<>();
TADManager tadManager = Nd4j.getExecutioner().getTADManager();
val ptrMap = new IntPointer(map);
long[] ptrs = new long[arrays.size()];
for (int i = 0; i < arrays.size(); i++) {
INDArray array = arrays.get(i);
Nd4j.getCompressor().autoDecompress(array);
int[] dimension = dimensions.size() > 1 ? dimensions.get(i) : dimensions.get(0);
Pair tadBuffers = tadManager.getTADOnlyShapeInfo(array, dimension);
list.add(tadBuffers);
Pointer hostTadShapeInfo = tadBuffers.getFirst().addressPointer();
DataBuffer offsets = tadBuffers.getSecond();
if (offsets.length() != numTads)
throw new ND4JIllegalStateException("Can't symmetrically shuffle arrays with non-equal number of TADs");
if (offsets == null)
throw new ND4JIllegalStateException("Offsets for shuffle can't be null");
dataPointers.put(i, array.data().addressPointer());
shapePointers.put(i, array.shapeInfoDataBuffer().addressPointer());
offsetPointers.put(i, offsets.addressPointer());
tadPointers.put(i, tadBuffers.getFirst().addressPointer());
}
if (Nd4j.dataType() == DataBuffer.Type.DOUBLE) {
nativeOps.shuffleDouble(dummy, dataPointers, shapePointers, dataPointers, shapePointers, arrays.size(),
ptrMap, tadPointers, offsetPointers);
} else if (Nd4j.dataType() == DataBuffer.Type.FLOAT) {
nativeOps.shuffleFloat(dummy, dataPointers, shapePointers, dataPointers, shapePointers, arrays.size(),
ptrMap, tadPointers, offsetPointers);
} else {
// HALFs
}
dataPointers.address();
shapePointers.address();
tadPointers.address();
offsetPointers.address();
}
/**
* This method converts Half-precision databuffer to current dType buffer.
*
* @param buffer
* @return
*/
/*
@Override
public DataBuffer restoreFromHalfs(DataBuffer buffer) {
if (buffer.dataType() != DataBuffer.Type.COMPRESSED)
throw new IllegalStateException("DataBuffer contains wrong data: " + buffer.dataType());
CompressedDataBuffer comp = (CompressedDataBuffer) buffer;
CompressionDescriptor descriptor = comp.getCompressionDescriptor();
DataBuffer targetBuffer = Nd4j.createBuffer(descriptor.getCompressedLength() / 2);
if (Nd4j.dataType() == DataBuffer.Type.DOUBLE) {
nativeOps.convertHalfsToDoubles(
null,
comp.addressPointer(),
(int) descriptor.getCompressedLength() / 2,
targetBuffer.addressPointer()
);
} else if (Nd4j.dataType() == DataBuffer.Type.FLOAT) {
nativeOps.convertHalfsToFloats(
null,
comp.addressPointer(),
(int) descriptor.getCompressedLength() / 2,
targetBuffer.addressPointer()
);
} else {
throw new UnsupportedOperationException("Target dtype isn't supported: " + Nd4j.dataType());
}
return targetBuffer;
}
*/
/**
* This method converts Single/Double precision databuffer to Half-precision databuffer
*
* @param buffer
* @return
*/
/*@Override
public DataBuffer convertToHalfs(DataBuffer buffer) {
// we allocate pointer
ShortPointer pointer = new ShortPointer(buffer.length());
if (buffer.dataType() == DataBuffer.Type.DOUBLE) {
nativeOps.convertDoublesToHalfs(
null,
buffer.addressPointer(),
(int) buffer.length(),
pointer
);
} else if (buffer.dataType() == DataBuffer.Type.FLOAT) {
nativeOps.convertFloatsToHalfs(
null,
buffer.addressPointer(),
(int) buffer.length(),
pointer
);
} else {
throw new UnsupportedOperationException("Source dtype isn't supported: " + buffer.dataType());
}
CompressionDescriptor descriptor = new CompressionDescriptor(buffer, new Float16());
descriptor.setCompressedLength(buffer.length() * 2);
CompressedDataBuffer result = new CompressedDataBuffer(pointer, descriptor);
return result;
}
*/
/**
* This method converts Single/Double precision databuffer to Half-precision databuffer
*
* @param typeSrc
* @param source
* @param typeDst @return
*/
@Override
public INDArray convertDataEx(DataBuffer.TypeEx typeSrc, INDArray source, DataBuffer.TypeEx typeDst) {
if (source.isView())
throw new UnsupportedOperationException("Impossible to compress View. Consider using dup() before. ");
DataBuffer buffer = convertDataEx(typeSrc, source.data(), typeDst);
source.setData(buffer);
if (buffer instanceof CompressedDataBuffer)
source.markAsCompressed(true);
else
source.markAsCompressed(false);
return source;
}
@Override
public DataBuffer convertDataEx(DataBuffer.TypeEx typeSrc, DataBuffer source, DataBuffer.TypeEx typeDst) {
int elementSize = 0;
if (typeDst.ordinal() <= 2)
elementSize = 1;
else if (typeDst.ordinal() <= 5)
elementSize = 2;
else if (typeDst.ordinal() == 6)
elementSize = 4;
else if (typeDst.ordinal() == 7)
elementSize = 8;
else
throw new UnsupportedOperationException("Unknown target TypeEx: " + typeDst.name());
DataBuffer buffer = null;
if (CompressionUtils.goingToCompress(typeSrc, typeDst)) {
// all types below 6 are compression modes
BytePointer pointer = new BytePointer(source.length() * elementSize);
CompressionDescriptor descriptor = new CompressionDescriptor(source, typeDst.name());
descriptor.setCompressionType(CompressionType.LOSSY);
descriptor.setCompressedLength(source.length() * elementSize);
buffer = new CompressedDataBuffer(pointer, descriptor);
} else {
CompressedDataBuffer compressed = (CompressedDataBuffer) source;
CompressionDescriptor descriptor = compressed.getCompressionDescriptor();
// decompression mode
buffer = Nd4j.createBuffer(descriptor.getNumberOfElements(), true);
}
convertDataEx(typeSrc, source, typeDst, buffer);
return buffer;
}
@Override
public void convertDataEx(DataBuffer.TypeEx typeSrc, Pointer source, DataBuffer.TypeEx typeDst, Pointer target,
long length) {
nativeOps.convertTypes(null, typeSrc.ordinal(), source, length, typeDst.ordinal(), target);
}
@Override
public void convertDataEx(DataBuffer.TypeEx typeSrc, Pointer source, DataBuffer.TypeEx typeDst, DataBuffer buffer) {
convertDataEx(typeSrc, source, typeDst, buffer.addressPointer(), buffer.length());
}
@Override
public void convertDataEx(DataBuffer.TypeEx typeSrc, DataBuffer source, DataBuffer.TypeEx typeDst,
DataBuffer target) {
convertDataEx(typeSrc, source.addressPointer(), typeDst, target.addressPointer(), target.length());
}
@Override
public INDArray createSparseCSR(double[] data, int[] columns, int[] pointerB, int[] pointerE, long[] shape) {
throw new UnsupportedOperationException();
}
@Override
public INDArray createSparseCSR(float[] data, int[] columns, int[] pointerB, int[] pointerE, long[] shape) {
throw new UnsupportedOperationException();
}
@Override
public INDArray createSparseCSR(DataBuffer data, int[] columns, int[] pointerB, int[] pointerE, long[] shape) {
throw new UnsupportedOperationException();
}
@Override
public INDArray createSparseCOO(double[] values, int[][] indices, long[] shape) {
throw new UnsupportedOperationException();
}
@Override
public INDArray createSparseCOO(float[] values, int[][] indices, long[] shape) {
throw new UnsupportedOperationException();
}
@Override
public INDArray createSparseCOO(double[] values, long[][] indices, long[] shape) {
throw new UnsupportedOperationException();
}
@Override
public INDArray createSparseCOO(float[] values, long[][] indices, long[] shape) {
throw new UnsupportedOperationException();
}
@Override
public INDArray createSparseCOO(DataBuffer values, DataBuffer indices, long[] shape) {
throw new UnsupportedOperationException();
}
@Override
public INDArray createSparseCOO(DataBuffer values, DataBuffer indices, DataBuffer sparseInformation, long[] shape) {
throw new UnsupportedOperationException();
}
@Override
public INDArray createSparseCOO(DataBuffer values, DataBuffer indices, long[] sparseOffsets, int[] flags, int[] hiddenDimensions, int underlyingRank, long[] shape) {
throw new UnsupportedOperationException();
}
@Override
public INDArray sort(INDArray x, boolean descending) {
if (x.isScalar())
return x;
if (x.data().dataType() == DataBuffer.Type.FLOAT) {
NativeOpsHolder.getInstance().getDeviceNativeOps().sortFloat(null, (FloatPointer) x.data().addressPointer(), (LongPointer) x.shapeInfoDataBuffer().addressPointer(), descending);
} else if (x.data().dataType() == DataBuffer.Type.DOUBLE) {
NativeOpsHolder.getInstance().getDeviceNativeOps().sortDouble(null, (DoublePointer) x.data().addressPointer(), (LongPointer) x.shapeInfoDataBuffer().addressPointer(), descending);
} else {
throw new UnsupportedOperationException("Unknown dataype " + x.data().dataType());
}
return x;
}
@Override
public INDArray sort(INDArray x, boolean descending, int... dimension) {
if (x.isScalar())
return x;
Arrays.sort(dimension);
Pair tadBuffers = Nd4j.getExecutioner().getTADManager().getTADOnlyShapeInfo(x, dimension);
if (x.data().dataType() == DataBuffer.Type.FLOAT) {
NativeOpsHolder.getInstance().getDeviceNativeOps().sortTadFloat(null,
(FloatPointer) x.data().addressPointer(),
(LongPointer) x.shapeInfoDataBuffer().addressPointer(),
(IntPointer) Nd4j.getConstantHandler().getConstantBuffer(dimension).addressPointer(),
dimension.length,
(LongPointer) tadBuffers.getFirst().addressPointer(),
new LongPointerWrapper(tadBuffers.getSecond().addressPointer()),
descending);
} else if (x.data().dataType() == DataBuffer.Type.DOUBLE) {
NativeOpsHolder.getInstance().getDeviceNativeOps().sortTadDouble(null,
(DoublePointer) x.data().addressPointer(),
(LongPointer) x.shapeInfoDataBuffer().addressPointer(),
(IntPointer) Nd4j.getConstantHandler().getConstantBuffer(dimension).addressPointer(),
dimension.length,
(LongPointer) tadBuffers.getFirst().addressPointer(),
new LongPointerWrapper(tadBuffers.getSecond().addressPointer()),
descending);
} else {
throw new UnsupportedOperationException("Unknown dataype " + x.data().dataType());
}
return x;
}
@Override
public INDArray sortCooIndices(INDArray x) {
throw new UnsupportedOperationException("Not an COO ndarray");
}
}
